Enterprise IIoT and Analytics in the Process Industriesweb.stanford.edu/class/ee392b/lecture/may28/OSIsoft.pdf · Title: OSIsoft Corporate PowerPoint Template Author: Meredith Picerno

© Copyright 2019 OSIsoft, LLC© Copyright 2019 OSIsoft, LLC

May 28, 2019

John Matranga, Director of Customer Innovation and Academia, OSIsoft

Ahmad Fattahi, Manager of Data Science Enablement, OSIsoft

Enterprise IIoT and Analytics in the Process Industries

© Copyright 2019 OSIsoft, LLC


OSIsoft is Trusted by the World’s Leading Companies

© Copyright 2019 OSIsoft, LLC 4


https://www.google.com/url?sa=i&rct=j&q=&esrc=s&source=images&cd=&cad=rja&uact=8&ved=0ahUKEwj-mPK79ODTAhVEImMKHQCJB4sQjRwIBw&url=https://techcrunch.com/2015/04/29/microsoft-announces-azure-data-lake-a-data-repository-for-big-data-analytics/&psig=AFQjCNHCCU6wxRecx8tThVF6CMZoDxE-oQ&ust=1494353779979852



Early Control Room – Circa 1960’s

• Central Control Room • Hundreds of sensors• Silos of Signals• Analog Readouts• Home Runs Needed for

Piping and Wires• Integration at the Human• Simple Control

Picture Source: en.wikipedia.org/wiki/Distributed_control_system


Distributed Control System (DCS)

• Introduced Late 70’s• Thousands of sensors• Rolled Out In the 80’s/90’s• Updated Slightly Over Time• DSC Allows for Plant-wide

Control • Advanced, Multivariable

Control• Typically multiple DCS in a

plant (7 in HI Independent)• VAX >> Unix >> Windows

Picture Source: en.wikipedia.org/wiki/Distributed_control_system


Electrical Infrastructure Supply Chain - Today

Source: California Independent System Operator (CALISO)

Redundant, Distributed Control CenterGeography Dispersed/Diverse Control

Cross Company Business Control(Circa 1990’s - Today)



The reality of new sensors today

1° Network

2° Network


“If somebody screws up yourhome thermostat,

nobody dies”- Plant Operations Manager


Reality of data science today

Insights

Operations

Expertise

DataScientists

AI

IT


Data and 1st Principles Physics

14

Data DrivenAnalytics

Physics DrivenAnalyticsHow to Best Mix ??


Real-World Examplesfrom Process Industries

15


Oil and GasReal-time Estimations and Online Learning for Industrial Assets at Total - Total


Oil and Gas


Oil and Gas


Oil and Gas


Food and BeverageBrewing Beer the Smart Way – Deschutes Brewery


Food and Beverage


Food and Beverage


Food and Beverage

Modernized Data Management – Historical, Real-Time, and Predictive


Food and Beverage


Power GenerationOptimizing Predictive Maintenance by Integrating Vibration Data and Process DataSilicon Valley Power


Power Generation


Power Generation


Power Generation


Principles of Analytics forProcess Industries

30


You need a Solid Business Question

31

Focus on Business Value• Sharp• Supported by existing or feasible data

Then start brainstorming ideas• Open to a large volume of ideas

Narrow down• Include SMEs and data professionals

Start building• Build KPI measurements in the process

Constantly validate with SMEs


Different Personas Have Different Goals

32

Business Goal Analytics GoalReduce the cost of production Identify what variable contributes

most to scrap material

Increase profit margin from Unit A Optimize the choice input material based on market price/demand

Improve the quality of our hires Identify success measures and translate into indicators on resumes

Make us safer against cyber attacks Identify anomalies in our network traffic beyond significance threshold


control? predict?

Is the goal of the project to…


Why Model Explainability Matters?

34

Winning trust for a critical system

Regulatory reasons

Improve chances of approval

Post-mortem analysis in case of failure is crucial


The Explainability Tradeoff

Source: ResearchGate GmbH


Computational Complexity Is Important


Data Engineering as Important as Data Science

• Development is different from Production• Data Governance challenges• Latency• Regulatory considerations• Engineering cost matters


Some Soft Factors

Balancing doubters and overly-zealous personas can be tricky

You need Subject Matter Experts• Features often need to be engineered from raw data

Fundamental challenges in mature companies• Eric Colson of Stitch Fix’s opinion• Focus on value - even incremental change can be very valuable• Look for the real problems and solve them• Look for (powerful) allies


“Data makes people think, emotions make them act.”- Antonio Damasio

Image credit: www.mycustomer.com

Tell Stories!


Make the Process Reproducible

CRISP-DM



Plant & Enterprise

42

Assets

MultipleSensors

Plant

MultipleAssets

Enterprise

MultiplePlants

PI System


The Digital Transformation


The Digital Transformation New Opportunity


Enterprise and Community scenarios have unique opportunities and challenges

45

Assets

MultipleSensors

Plant

MultipleAssets

Enterprise

MultiplePlants

Community

MultipleEnterprises

Sensors

Millions of Smart Devices

PI System

PDC/Edge

OSIsoft Cloud Services


OCS: Complementing the PI System

Remote Operations Monitoring

Connected Community

Data Science Enablement

PaaS & Partner App.

Enterprise Data

OSIsoftCloud Services

AutomationAssets

cloud.osisoft.com

Edge DataEdge Devices

Edge DataEdge Devices


• Over 1,200 Customer Talks• Video and Slides• Company and Industry Issues

Resources – www.OSIsoft.com


John [email protected] of Customer Innovation and AcademiaOSIsoft, LLC

48

Ahmad [email protected] of Data Science EnablementOSIsoft, LLC

mailto:[email protected]

mailto:[email protected]


Enterprise IIoT and Analytics in the Process Industriesweb.stanford.edu/class/ee392b/lecture/may28/OSIsoft.pdf · Title: OSIsoft Corporate PowerPoint Template Author: Meredith Picerno

Documents